Segmentation of fMRI based on graph space construction

The functional magnetic resonance imaging (fMRI) is an advanced medical imaging technique based on blood oxygen level dependence (BOLD), which has the higher time and the spatial resolution. Because the BOLD-fMRI signal changes only about 0.5-2%, how to examine and locate the functional activation signal from those selected images with low signal to noise ratio accurately and reliably is the open question. In this paper, based on phase space construction, the proposed activation lags between pictures are estimated by the minimum mutual information principle. Then, the image as the activation reachs to stability is obtained for further clustering analysis. Finally, an image segmentation algorithm based on pulse-coupled neural network (PCNN) is presented in this paper. PCNN dynamically evaluates similarity between any two samples owing to the outstanding centralization characteristic based on the vicinity in space and the comparability of brightness. It has higher accuracy and faster performance than those classical clustering algorithms. Experimental results with fMRI images have shown the effectiveness of the proposed algorithm.